Occupational Polarization

Definition

Occupational polarization is the secular shift in the distribution of employment away from middle-skill, middle-wage occupations and toward the tails: high-skill, high-wage professional/technical/managerial jobs on one end, and low-skill, low-wage service and laborer jobs on the other. The "barbell" shape of employment change — gains at top and bottom, losses in the middle — has been documented across the U.S., United Kingdom (UK), Germany, and most other industrialized economies from the 1970s forward. The core mechanism is routine-biased technological change: automation substitutes for the routine cognitive and manual tasks concentrated in middle-skill clerical, administrative, and production occupations, while complementing the abstract reasoning of high-skill workers and leaving the manual dexterity/interpersonal content of low-skill service jobs largely untouched.

Key Ideas

Three-tier occupational structure: Low-skill (services, transport, construction, laborers); middle-skill (production, clerical, admin, sales); high-skill (professional, technical, managerial). Polarization = middle hollows out, top and bottom expand.
Magnitude of middle-skill collapse: Middle-skill occupations fell from $57.3\%$ of employment in 1979 to $48.6\%$ in 2007 and $45.7\%$ by 2009 (Autor 2010). This is not a post-crisis artifact — it predates the Great Recession by decades.
Polarizing recession: Within-recession occupational employment growth ( $2007$ – $2009$ ) correlates $0.76$ with pre-recession growth ( $2001$ – $2007$ ) — the Great Recession accelerated ongoing polarization rather than creating a new pattern (Autor 2010).
International breadth: All 16 European Union (EU) countries (EU-15 + Norway) show employment polarization $1993$ – $2006$ (Autor 2010; Goos, Manning, and Salomons 2014). The pattern is not U.S.-specific.
Institutional mediation — not all countries fully polarize (Oesch and Rodriguez Menes 2010): Four-country comparison (Britain, Germany, Spain, Switzerland 1990–2008) finds universal top-end upgrading and middle collapse, but bottom-quintile expansion only in Britain (+3.1%) and Spain (+3.8%) — not in Germany (−2.5%) or Switzerland (−1.8%). The divergence is explained by wage-setting institutions: where compressed wages raise the relative price of low-paid interpersonal service labor (Germany Q1 median = 78% of overall median; Switzerland = 73%), such jobs are not created. Britain (65%) and Spain (68%) allow creation of low-paid service jobs, producing actual polarization. Routinization thus produces "polarized upgrading" in flexible-wage economies but closer to pure upgrading in compressed-wage economies. See Oesch and Rodriguez 2010 — Upgrading or Polarization.
Immigration-driven bottom-end expansion: In Britain and Spain, all of the Q1 employment expansion was attributable to immigrant workers downgrading upon arrival (language barriers, credential non-recognition). Native workers declined in Q1 in both countries. Without immigration, Britain would have experienced a ~3 pp decline in Q1 — indistinguishable from Germany (Oesch and Rodriguez 2010). This implies that occupational "polarization" in flexible-wage economies partly reflects immigration supply shocks, not only demand-side routinization.
Asymmetric impact by education: College workers have shifted modestly from middle to both top and bottom; non-college workers have shifted almost entirely from middle to bottom (low-skill). Among non-college workers, $12.3$ percentage points (pp) of employment moved from middle-skill to low-skill occupations between $1980$ and $2016$ (Autor 2019).
Gender asymmetry: Female workers shifted upward in the skill distribution across the full period; less-educated males shifted primarily downward into low-skill service work (Autor 2010).
College wage premium surge: College/high school (HS) hourly wage ratio rose from $\approx 1.50$ (1963) to $1.95$ (2009), entirely post-1980 (Autor 2010). Non-college male real wages fell $-12\%$ (HS graduates) and $-16\%$ (HS dropouts) from $1979$ – $2007$ .
Urban concentration: Pre-1970, dense cities offered non-college workers disproportionate access to middle-skill production and clerical jobs (an urban occupational skill gradient of $\approx 25$ pp). As those jobs disappeared, polarization fell more heavily on urban labor markets. By 2015, the urban middle-skill gradient for non-college workers had completely reversed.
Wage consequences: The collapse of middle-skill urban work eroded the non-college urban wage premium steeply after 1990, contributing to falling real wages for non-college workers ( $10$ – $20$ log points below 1980 levels by 2018).
Both automation and trade: Urban manufacturing decline and international trade competition (especially from China, post-2000) drove production job losses; office computing drove the parallel collapse of clerical and administrative jobs.

How It Works

The standard task-based account (Autor, Levy, Murnane 2003; Autor and Dorn 2013) distinguishes routine vs. non-routine tasks and cognitive vs. manual tasks:

Task type	Computerizable?	Trend
Routine cognitive (clerical, data entry)	Yes	Declining employment
Routine manual (assembly, production)	Largely yes	Declining employment
Non-routine cognitive (analysis, management)	No (complement to capital)	Rising employment
Non-routine manual (service, personal care)	No (context-dependent)	Stable/rising employment

Middle-skill jobs are intensive in routine tasks; polarization is the labor-market implication of automating them. At the top, automation complements educated workers who manage and design systems. At the bottom, tasks requiring physical presence and interpersonal interaction resist automation.

The geographic mechanism (Autor 2019): Middle-skill jobs were co-located with high-skill knowledge workers in dense cities because they supported production and office functions. When automation eliminated those middle-skill functions, it preferentially reduced urban (not rural) employment for non-college workers. Non-college workers in cities were displaced downward into the same low-skill service jobs that rural non-college workers always held — eliminating the urban occupational wage gradient.

Why It Matters

Non-college real wages: The partial-equilibrium accounting in Autor (2019) can explain essentially all of the observed fall in high school dropout and graduate wages $1970$ – $2015$ , once geography of polarization is included.
DI applications: The displacement of non-college workers from stable, middle-skill jobs is a supply-side driver of disability insurance (DI) applications. Job loss from trade-exposed manufacturing (see China Trade Shock) and automation increases DI application rates — these are the same workers who used to hold urban middle-skill jobs.
Marriage market and family structure: Loss of stable middle-skill manufacturing jobs reduces male earnings and marriageability, driving nonmarital birth rates up (see Marriage Market).
Geographic divergence: Dense cities have become more college-intensive while simultaneously offering worse opportunities for non-college workers, producing spatial sorting and reducing non-college in-migration to high-wage cities.

Open Questions

Will new occupational categories ("wealth work," "frontier jobs," "last mile" jobs) eventually generate a new tier of middle-skill work for non-college workers, or does the bifurcation persist?
What is the role of the "fissuring of the workplace" (Weil 2014) — outsourcing of support functions from core firms — in amplifying polarization beyond what automation predicts?
To what extent does the declining non-college urban wage premium reflect lower wages for the same workers versus compositional shifts in which non-college workers remain in cities?
Does the reversal of the urban non-college wage gradient reduce incentives for non-college migration to cities, and if so, does that eventually equalize low-skill wages across geographies?